human is the awkward prepubescent phase between animal and machine

blog:personal

25719 0402

Neuralink pārliecinoši rāda priekšzīmi [1][2] tam, kā jābūvē smadzeņu-datoru interfeisus. Publikācija apraksta tikai agrīnu koncepta demonstratoru ar virkni skaidri iezīmētu tālāku uzlabojumu, bet jau tas vien ir daudz iespaidīgāks par visiem iepriekšējiem lielo projektu gala rezultātiem, kopā ņemtiem (pie tam ar tikai nepilnu 3 gadu ilgu R&D fāzi).

We have built arrays of small and flexible electrode “threads”, with as many as 3,072 electrodes per array distributed across 96 threads. We have also built a neurosurgical robot capable of inserting six threads (192 electrodes) per minute. Each thread can be individually inserted into the brain with micron precision for avoidance of surface vasculature and targeting specific brain regions. The electrode array is packaged into a small implantable device that contains custom chips for low-power on-board amplification and digitization.

... tālāk ... )

[1] https://www.biorxiv.org/content/10.1101/703801v1
[2] https://www.youtube.com/watch?v=r-vbh3t7WVI


tags: , , , ,
state: blank
sound: Ager Sonus - Mithra: Banished

261118 1409

Don’t listen to the bioethicists, talk to the Chinese.


tags: , , , ,
state: working
sound: Atrium Carceri & Eldar - Sacrosanct: Sol

311018 1754

The concern about uFAI as an existential threat is relevant but we shouldn’t think it implies it would take no less than uFAI to obsolete us. There is already almost nothing we are able to do unequivocally better than bespoke narrow AI systems available today. Happy Halloween.


tags: , , , ,
state: working
sound: Dahlia's Tear - Through The Nightfall Grandeur: Lost In The Crystalline Enigma

15218 0311

Ir iznācis pirmais 800 lpp. sējums [1] “Longevity Industry Landscape Overview 2017” [2] apskatā. Laika trūkuma un apjoma (un vismaz daļējas aprioras tēmas pazīstamības) dēļ droši vien šobrīd to tikai pāršķirstu kopā ar kopsavilkumu [3] un detalizētu analīzi atstāšu vēlākam laikam, bet kopaina ir skaidra: pirmo reizi zinātne un medicīna pret novecošanas procesu kā tādu (pretstatā atsevišķām to pavadošām vai no tā izrietošām patoloģijām) sāk attiekties kā pret novēršamu un ārstējamu parādību, un nepilnas desmitgades laikā šīs pūles no mazām konferencēm un vēstulēm ziņu grupās ir izaugušas par masīvu multidisciplināru R&D metaprojektu, kur gadā tiek ieguldīti miljardiem dolāru un miljoniem cilvēkstundu gan privātās iniciatīvās, gan valstu institūtos un laboratorijās visā pasaulē.

Pagaidām zināmā mērā pietrūkst koordinācijas un komunikācijas dažādo projektu un institūtu starpā, taču, ņemot vērā pasākuma apjomu un tā neilgo vēsturi un velkot paralēles ar attīstības gaitu citās jomās, prognozēju, ka tā ir pārejoša parādība.

[1] http://lir.website/pdfdata/pdf4.pdf
[2] http://longevity.international/longevity-industry-landscape-overview-2017
[3] https://singularityhub.com/2018/02/14/we-read-this-800-page-report-on-longevity-research-so-you-dont-have-to/


tags: , , , , , ,
state: working
sound: Randal Collier-Ford - Remnants: Horns Of Eden

17118 2115

The crossover between machine learning and CRISPR [1][2][3] was as inevitable as it will be impactful in the near future. Kind of surprising to see Microsoft pioneering this, though.

[1] https://gizmodo.com/microsoft-wants-to-use-artificial-intelligence-to-make-1821956922
[2] https://www.nature.com/articles/s41551-017-0178-6
[3] https://crispr.ml/


tags: , , , , , ,
state: working
sound: Cryobiosis - Inner Stasis: Permanent Void

7118 1149

Pessimistic: CRISPR probably won't work well for IQ enhancement because the inheritable component of IQ is massively polygenic, likely encompassing tens of thousands of SNPs to predict only several percentage points of variability. only a few of those SNPs have enough individual impact to bother, worth a couple IQ points total. Selecting an initial batch of embryos based on single-cell sequencing and running the results through a GWAS model for IQ, then using those few specific high-value loci to set up parameters for either CRISPR or iterated embryo selection [1][2] could be a much more practical, if also much slower, approach.

Optimistic: IQ is massively polygenic as above but the relevant SNPs form clusters with directly traceable and quantifiable impact on the microstructure of the brain, responsible for much (up to 40 % and possibly more) of the observed inherited variability [3]. These clusters could be used to directly narrow down the target set for CRISPR.

[1] https://wiki.lesswrong.com/wiki/Iterated_embryo_selection
[2] http://onlinelibrary.wiley.com/wol1/doi/10.1111/1758-5899.12123/full
[3] http://www.cell.com/neuron/fulltext/S0896-6273(17)31092-9


tags: , , , ,
state: thoughtful
sound: Cor Serpentis - Unterchthonia: Artifact

131117 0319

Optiskie neiromorfie čipi. Atjaunoju uz augšu savu varbūtību tam, ka brīdī, kad tiks sasniegts cilvēka WBE, būs pieejama arī par standarta eksaflopu mēroga HPC sistēmām kompaktāka un jaudīgāka (uz specializācijas rēķina un pateicoties programmatūras līmeņa abstrakciju iztrūkumam) aparatūra.

[1] http://advances.sciencemag.org/content/3/9/e1700160.full
[2] https://arxiv.org/abs/1611.02272


tags: , , , , ,
state: awake
sound: Forma Tadre - Navigator: Date Unknown

9917 2329

Prediction: Given enough time, blockchains will out-compete and replace most other forms of networks.

Humans are a highly networked species, the first one to network across genetic and geographic boundaries and thus expand its dominion. Networks allow humans to cooperate in an organised manner and to allocate the fruits of the cooperation. Overlapping networks create and organize our society, being the physical, mental, and digital roads connecting us all. Money is a network. Religion is a network. A corporation is a network. Travel and trade routes form a network. Electricity is a network. Of course, the Internet is a network. Ultimately, civilisation and culture comprises a network of such networks.

Networks must be organized according to rules. They require governance to enforce these rules against cheaters and freeloaders. Networks have “network effects”. Adding a new participant increases the value of the network for all existing participants. Network effects thus create a winner-take-all dynamic. The leading network tends towards becoming the only network. The rulers of these networks become the most powerful people in the society. Some networks are run by kings and priests who choose what is money and law, what is sacred and profane. Rule is closed to outsiders and based on power. Many are run by corporations – the social network, the search network, the phone or cable network; closed but (at least initially) meritocratic. Some are run by elites – the university network, the medical network, the banking network; somewhat open and somewhat meritocratic. A few are run by the mob – democracy, most of the current Internet, the commons; open, but not meritocratic, and, generally, comparatively very inefficient.

The XX century created a new kind of network – market networks; open, efficient, and meritocratic. Merit in markets is determined by the commitment of resources. The typical resource-equivalent is, of course, money, which itself is just a form of reified time. Market networks work where there is a commitment of money, otherwise they are just mob networks. Their applications have been somewhat limited, until now. Blockchain is the new invention that allows meritorious participants in an open network to govern without a ruler and without conventional money. They are merit-based, tamper-proof, open systems. The meritorious are those who work to advance the network. As society gives you money for giving society what it wants, blockchains give you coins for giving the network what it wants. It’s important to note that blockchains pay in their own coin, not the common money of financial markets. Blockchains pay in coin, but the coin itself just tracks and quantifies the amount of work done, and different blockchains demand different work: Bitcoin pays for securing the ledger, Ethereum pays for (executing and verifying) computations, etc. Blockchains combine the openness of the Internet with the merit-based modus operandi of markets.

potential blockchain landscape )

To a blockchain, merit can mean security, computation, prediction, attention, bandwidth, power, storage, distribution, content, and so forth. Blockchains port the market model into places where it couldn’t go before. Blockchains’ open and merit–based markets can replace networks previously run by kings, corporations, aristocracies, and mobs with networks governed by anyone with merit to the network, optionally but preferentially delegating arbitrary complex parts of the governance process to formally verifiable smart contracts co-approved by the meritorious users of the network. It’s nonsensical to have a blockchain without a coin just like it’s nonsensical to have a market without money. It’s also nonsensical to have a blockchain controlled by a sovereign, a corporation, an elite, or a mob. Blockchains, therefore, provide novel ways to govern networks, whether for banking, research, voting, law, search, social media, phone and energy grids, or computation and communication in general.


tags: , , , , , ,
state: awake
sound: Mentallo & The Fixer - Burnt Beyond Recognition: Radiant

18517 1556

Ja šo divu papīru secinājumi izrādīsies atbilstoši realitātei, tad vienā nedēļā būs likti stabili pamati > 95% Visumu sastādošā “stuffa” (attiecīgi – tumšās enerģijas un tumšās matērijas) saprašanai, pie reizes atrisinot plašā publikā mazāk zināmo, bet fizikā ne mazāk aktuālo problēmu ar teorētisko modeļu prognozētās (gigantiska) un praksē novērotās (ārkārtīgi neliela) vakuuma enerģijas vērtības neatbilstību. Exciting times, sekosim līdzi.

tags: , , , ,
state: awake
sound: Neurotech - The Decipher Volumes: Below These Scars

22317 2114

Padomāju par šo diskusiju. Daži secinājumi, neviens no tiem ļoti oriģināls:

  • Zinātnes metodoloģija ir apolitiska, bet zinātniskā darbība tāda nav, jo to veic cilvēki ar saviem uzskatiem, vērtībām, grupu piederībām u.c.tml. lietām, kas lielā[kajā?] daļā gadījumu netiek atstātas aiz laboratoriju durvīm;
    • Tajā pašā laikā, virzība “pēc iespējas mazāk politikas” virzienā ir deziderāts gan zinātnē, gan citur;
    • Kas attiecas uz ētikas jautājumiem tajās jomās, kur tie ir aktuāli un relevanti, tad uzskatu, ka būtu vēlams, pirmkārt, šos jautājumus izlemjošos kodeksus padarīt more streamlined, savukārt pašu izlemšanas procesu atstāt profesionālām ētikas komisijām, kuru darbs notiek attiecīgās zinātnes disciplīnas pārstāvjiem caurspīdīgā veidā;
  • Zinātniekiem vajadzētu būt responsible par pētījumu un to rezultātu kvalitāti, bet ne culpable par šo pētījumu rezultātu iespējamiem nevēlamiem pielietojumiem further down the road;
    • To var uztvert kā motivāciju tiekties iegūt pēc iespējas lielāku suverenitāti laboratorijām, institūtiem un R&D kompānijām (by extension – kompāniju R&D interesēm in general);
  • As a fully general argument – aktīvisms lielākoties ir nevēlams, informēšana lielākoties ir vēlama, savukārt robeža starp vienu un otru nereti ir novelkama ar vērā ņemamām grūtībām;
    • Mans arguments par aktīvisma nevēlamību izriet gk. no a.m. “pēc iespējas mazāk politikas” argumenta un tā pamatojuma, kā arī maniem no iepriekšējā izrietošajiem uzskatiem par to, ka tas politikas daudzums, no kura nav iespējams praktiski izvairīties, būtu atstājams profesionālu (t.i., eksplicīti un formāli amatu ieņemošu) statesmanship ekspertu pārziņā;
  • Zinātne ir tikpat pakļauta optimizācijas un automatizācijas tendencei, kā visas pārējās lietas, and that’s OK; mēs vai nu pārbūvēsim sevi, lai turpinātu būt konkurētspējīgi, vai arī kļūsim par zemsvītras piezīmi mūsu ļoti citādo pēcteču hronikās (ja pat tik daudz).

Piezīme sev par vispārzināmo, bet pārāk bieži aizmirsto: pirms diskusijas censties novienoties par lietoto jēdzienu definīcijām un tos atpakot, lai ar pēc iespējas lielāku varbūtību izvairītos no tā, no kā sastāv vismaz trešdaļa minētās diskusijas, gan manā, gan, to the best of my ability to observe, citu izpildījumā.

Piezīme sev par nesaistītām vai meta lietām: pabeigt migrēt bloga saturu uz jaunajiem confidence un importance tagiem, šobrīd skripti kaut kādēļ turpina kārties.


tags: , , , , , ,
state: awake
sound: Aegri Somnia - Endtime Psalms: We Were Stardust
Powered by Sviesta Ciba